time to fish or cut bait?

[Sven]

Didn't Vas recently correspond with someone saying 20% of fruit was in rybka? Recent post ... too lazy to search it ...

Don't remember seeing that, but doesn't mean it didn't happen.

What is truly interesting is how the number changed from ZERO in the interview to now 20 % once Zak's report was released.

I think he's referring to (from the "My recent correspondence from Vasik Rajlich" thread):

(5) It seems to be guesswork. I'd estimate that 40% of the points are wrong, 40% are standard chess/computer chess concepts, and 20% are direct Fruit influence on Rybka.

As if Zach just posted random stuff that he hadn't researched at all....

That "someone" was me ...
Here is the correct link to my article.
The "20%" statement has been grossly misinterpreted by some people since then. Everyone can read easily that Vat meant "20% of the Zach points are direct Fruit influence" and not "20% of Fruit code are copied to Rybka". This is a very huge difference since the total number of issues mentioned by Zach on his "evidence" pages is about 10 or 11, so Vas' statement means that he thinks that about 2 or 3 of the Zach points are mentioning topics where Fruit had direct influence on Rybka 1.

And that "direct influence" is indeed subject of interpretation. I think it is clear for everyone what Vas meant here, though.

I kindly ask everyone to research carefully before publishing such apparently wrong statements.

Sven

[Sven]

Please be clear. Are you making any GPL violation claims against either engine? Ideas used from Fruit are not a violation unless code is taken outside of the GPL. I am sure the team of people that worked on Fruit 2.1 are proud to have impacted other engines in a positive way.

I am saying, unequivocally, that code from fruit was copied and used in Rybka 1. Whether or not this code remains in current versions is unknown, but highly probable. ip* is unknown. I have not personally compared it to fruit, but that might be an interesting comparison. Although I am not sure it would be convincing since if you copy from the original, or a copy of the original, you could end up with the same thing and it would be pretty tough to decide whether the thing is a copy of the original or the copy.

Because the souce code for Rybka 1 is not availble, the only thing that is available is Zach's reverse engineering job.

Since Rybka 1 is bitboard and not array based like Fruit, it is literally impossible to simply cut and paste things like eval and search.

Actually, search has nothing to do with bitboards. Just look at search.c in Crafty. Evaluate() is clearly a different animal, however. But there are large chunks of a chess engine that are independent of the board representation, yet they are important to playing chess. Move ordering/selection, search itself, hashing, extension/reduction stuff, and even some parts of eval that don't care about piece placement (material balance/imbalance, etc.)

There is no single evidence that Rybka 1 contains *search* code that has been taken from another engine. Quite the contrary, it is well-known that Rybka has a very unique search implementation. Zach "evidence" pages refer to evaluation only (which does not mean for me that the "eval" case is already clarified). So please stop trying to be suggestive of Rybka 1 containing search code from Fruit.

Here is what is literally and undeniably clear:
Vas carefully studied Fruit and then either copied code or wrote his own version. If he copied then he committed a crime. If he wrote his own version then he may have done somthing that chess programmers do not like but which would nonetheless be legal.

I think it is strange to say you know which of the two possibilities actually occurred.

Based on 40+ years of trying to prevent student plagiarism on homework assignments. One doesn't find an identical block of code here, an identical block of code there, in a program of any significant size (>20 lines of code). I could take the assignments from my X86 programming course, where programs 2, 3 and 4 are on the range of 30-50 lines each, and let you take assignment 2 from 20 students and compare them. You'd see what I mean. Very different approaches to produce the same result. Duplicate code just does not happen by accident.

You are suggestive again, making the readers believe that there were a proof of "duplicate code" from Fruit to Rybka 1. THERE ISN'T A PROOF !!

- Some PST values are similiar or identical (for the non-programmers: PST = piece square table = constant parameters for evaluation, most people agree that such constants are not covered by the GPL).

- Few small code snippets from evaluation, to be counted with few fingers of one hand, look similar but you can't tell whether code was copied or not.

- Finally, a couple of code snippets were presented by Zach as "Rybka code" which do not appear this way in Rybka (as Zach explicitly states by himself) but are an "equivalent" - so where's the point in these?

- Hundreds of other code parts from Fruit 2.1, which has about 10000 "netto" lines of source code (i.e., non-empty and non-comment), have not been referred to by Zach, most of them clearly being out of scope for "copying" due to their non-bitboard contents, and have definitely not influenced Rybka 1 other than by reuse of ideas.

What remains? UCI parser? Usage of strtok()? Open your eyes, Bob. You know that sentence about the "dead horse". Zach has done a good job but his analysis does *not* show what you want to make us all believe it does.

Sven

[Matthias Gemuh]

As if Zach just posted random stuff that he hadn't researched at all....

That "someone" was me ...
Here is the correct link to my article.
The "20%" statement has been grossly misinterpreted by some people since then. Everyone can read easily that Vat meant "20% of the Zach points are direct Fruit influence" and not "20% of Fruit code are copied to Rybka". This is a very huge difference since the total number of issues mentioned by Zach on his "evidence" pages is about 10 or 11, so Vas' statement means that he thinks that about 2 or 3 of the Zach points are mentioning topics where Fruit had direct influence on Rybka 1.

And that "direct influence" is indeed subject of interpretation. I think it is clear for everyone what Vas meant here, though.

I kindly ask everyone to research carefully before publishing such apparently wrong statements.

Sven

Vas probably did say "20% of the Zach points are direct Fruit influence", but isn't that a lie ?

Who needs Vas to understand the facts Zach presents ?

Far more accurate would be "100% of the Zach points are direct Fruit influence or copy/paste".

Matthias.

[Robert Flesher]

Didn't Vas recently correspond with someone saying 20% of fruit was in rybka? Recent post ... too lazy to search it ...

Don't remember seeing that, but doesn't mean it didn't happen.

What is truly interesting is how the number changed from ZERO in the interview to now 20 % once Zak's report was released.

I think he's referring to (from the "My recent correspondence from Vasik Rajlich" thread):

(5) It seems to be guesswork. I'd estimate that 40% of the points are wrong, 40% are standard chess/computer chess concepts, and 20% are direct Fruit influence on Rybka.

As if Zach just posted random stuff that he hadn't researched at all....

Peter[/quote]


No I was referring to the video interview with Vas when he states that Fruit played no roll in Rybka's strength or sudden appearance. ( this is not verbatim, however, i was his response) However, it changed to 20%, and this is shortly after Zak's Fruit and Rybka 1.0 comparision. Capiche ?

[Zach Wegner]

There is no single evidence that Rybka 1 contains *search* code that has been taken from another engine. Quite the contrary, it is well-known that Rybka has a very unique search implementation. Zach "evidence" pages refer to evaluation only (which does not mean for me that the "eval" case is already clarified). So please stop trying to be suggestive of Rybka 1 containing search code from Fruit.

There absolutely is evidence, I only ever completed the pages on evaluation. And no, Rybka does not have a very unique search implementation.

[Rolf]

Didn't Vas recently correspond with someone saying 20% of fruit was in rybka? Recent post ... too lazy to search it ...

Don't remember seeing that, but doesn't mean it didn't happen.

What is truly interesting is how the number changed from ZERO in the interview to now 20 % once Zak's report was released.

I think he's referring to (from the "My recent correspondence from Vasik Rajlich" thread):

(5) It seems to be guesswork. I'd estimate that 40% of the points are wrong, 40% are standard chess/computer chess concepts, and 20% are direct Fruit influence on Rybka.

As if Zach just posted random stuff that he hadn't researched at all....

Peter

Hi.

Verbatim and under oath Vas confidently told me that he took 25% from fruit. After I had cured from my stroke then he told me, Rolf, you can tell it to the boys on CCC, they will collapse. Wait, tell them I took 99%, and let's see what happens. In real I didnt even take 3% but just for the fun of it. Let them show their faces...

That is what happened along my true memory record. But then my English isnt that good after all. So, take it with a grain of salt. I dont even know if I really spoke with Vas and not a double or clone. Still I wanted to give the best report I could about my travel to Hungary or Poland.

Reporter regards

[Sven]

There is no single evidence that Rybka 1 contains *search* code that has been taken from another engine. Quite the contrary, it is well-known that Rybka has a very unique search implementation. Zach "evidence" pages refer to evaluation only (which does not mean for me that the "eval" case is already clarified). So please stop trying to be suggestive of Rybka 1 containing search code from Fruit.

There absolutely is evidence, I only ever completed the pages on evaluation. And no, Rybka does not have a very unique search implementation.

Then show your proof. As long as it is on your hard disk only nobody else can see it.

You will have a hard time proving it, though.

Sven

[Roger Brown]

Then show your proof. As long as it is on your hard disk only nobody else can see it.

You will have a hard time proving it, though.

Sven

Hello Sven,

Sounds like there is a decision about what will be shown before it is shown.

How would you know about the difficulty of the task without the attempt being made? I mean, Zach may have compelling evidence on his hard drive.

I hope that there is not an unscientific faith in things not seen...



Later.

[Milos]

Then show your proof. As long as it is on your hard disk only nobody else can see it.

You will have a hard time proving it, though.

Seams you don't believe Zach's word, ergo, you think he's a liar.
Seams in the same time you believe Vas's word (and not ask any additional proof from him), ergo, that's hypocrisy by definition.

[bob]

There is no single evidence that Rybka 1 contains *search* code that has been taken from another engine. Quite the contrary, it is well-known that Rybka has a very unique search implementation. Zach "evidence" pages refer to evaluation only (which does not mean for me that the "eval" case is already clarified). So please stop trying to be suggestive of Rybka 1 containing search code from Fruit.

There absolutely is evidence, I only ever completed the pages on evaluation. And no, Rybka does not have a very unique search implementation.

Then show your proof. As long as it is on your hard disk only nobody else can see it.

You will have a hard time proving it, though.

Sven

That is true. For _some_, no amount of "proof" will be enough.